ARTS: accurate recognition of transcription starts in human

by Sören Sonnenburg, Alexander Zien and Gunnar Rätsch

We develop new methods for finding transcription start sites (TSS) of RNA
Polymerase II binding genes in genomic DNA sequences. Employing Support Vector
Machines with advanced sequence kernels, we achieve drastically higher
prediction accuracies than state-of-the-art methods.

This is the stand-a-lone tool 'arts' for recognition of *human* tss. It
requires the shogun toolbox version at least 0.7.3 available from
http://www.shogun-toolbox.org and will create point predictions for each
nucleotide in the whole genome.


USAGE
=====

./arts [options] seq.fa

  -o OUTFILE, --outfile=OUTFILE
			  File to write the results to

If no option is specified arts will print results to stdout.

NOTE
====
*No* thresholding is done. In case you'd like ARTS to present positive
predictions with a very low false positive rate you will need to manually adjust
thresholds (which can be easily achieved by adding a constant bias term). 

Arts will output random results in regions containing lots of characters other
than A,C,G,T (e.g., N's).

Wiggle tracks, further information on the training data used can be found at
http://www.fml.tuebingen.mpg.de/raetsch/suppl/arts .

Variants of this system are used in our gene finder mgene (see
		http://www.mgene.org).
